The inference warm-up is performed 10 times, and the cycle is performed 100 times. Speed is tested on MNN 2.3.0 AArch64 with 2 threads by arm82 acceleration.Results of the mAP and speed are evaluated on COCO val2017 dataset, and the input resolution is the Size in the table.All checkpoints are trained with 400 epochs without distillation.From the perspective of model size and input image ratio, we have built a series of models on the mobile terminal to facilitate flexible applications in different scenarios. Precision is figured on models for 300 epochs.Speed is tested with TensorRT 8.4 on T4.Params and FLOPs of YOLOv6 are estimated on deployed models.Refer to Test speed tutorial to reproduce the speed results of YOLOv6.Speed is tested with TensorRT 7.2 on T4.Results of the mAP and speed are evaluated on COCO val2017 dataset with the input resolution of 640×640 for P5 models and 1280x1280 for P6 models.All checkpoints are trained with self-distillation except for YOLOv6-N6/S6 models trained to 300 epochs without distillation.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |